[lldb] Don't rely on ScriptInterpreterPythonImpl::Initialize in the unit tests (#82096)

The unit tests only test the Python objects and don't actually use
anything from the LLDB module. That means that all the additional
complexity in ScriptInterpreterPythonImpl::Initialize is overkill.

By doing the initialization by hand, we avoid the annoying
ModuleNotFoundError.

  Traceback (most recent call last):
    File "<string>", line 1, in <module>
  ModuleNotFoundError: No module named 'lldb'

The error is the result of us stubbing out the SWIG (specifically
`PyInit__lldb`) because we cannot link against libLLDB from the unit
tests.

The downside of doing the initialization manually is that we do lose a
bit of test coverage. For example, issue #70453 also manifested itself
in the unit tests.